vp-init

name	vp-init
description	Starts a Splunk custom visualization pack build. Gathers app name, brand context, tone, font strategy, data source, viz inventory, then routes to the right workflow.
when_to_use	Use when starting a new viz pack from scratch. Triggers on 'new viz pack', 'start viz project', 'build splunk vizs', 'custom visualization app', 'themed viz suite'.
disable-model-invocation	false
arguments	["app-name","brand"]
argument-hint	[app-name] [brand-description]

vp-init — start a viz pack build

Quick start

If arguments provided: /vp-init my_pack "Nike - bold, kinetic" Fill in what was given, then ask the remaining questions.

Otherwise, ask ALL 8 questions below. These feed directly into vp-design's design context — skipping any means the user gets asked twice.

1. Target format

Classic (Dashboard Studio + Simple XML) or Extension (Dashboard Studio only)?

• Classic — generates formatter.html + AMD visualization_source.js, packages with build_flat.js + tar. Works on all Splunk Enterprise 9.x+ and Splunk Cloud. This is the default.
• Extension — generates config.json + ESM visualization.js, packages with yarn. Requires Splunk 10.4+ with Dashboard Studio Extension API.

Default: Classic. If the user does not specify or is unsure, use Classic.

2. App name

Lowercase with underscores. Used as Splunk app ID and directory name. Example: nike_training_club, hospital_ops_viz

3. Brand / domain

Short description of brand identity AND industry/use case. Example: "Nike Training Club — bold, kinetic, volt on black, fitness domain"

4. Tone — 3 committable words

NOT "modern" or "clean" — those are dead categories. Push for specificity. Example: "kinetic, defiant, electric" or "clinical, precise, reassuring"

5. Font strategy

1-2 fonts max. Base64 embedded in viz CSS. System fonts = zero overhead. Example: "Barlow Condensed for display, monospace for values" or "system sans-serif only"

6. Data source

Demo CSV (recommended, default):

• Fastest path — viz works immediately after install
• Production SPL can be added later
• No Splunk access required during development

Production data:

• Requires Splunk MCP for data discovery
• Load spl-gotchas from splunk-spl before writing queries
• Recommended only if you already know your data schema

7. Viz inventory

Which vizs does this pack need? Can be specific names, general categories, or "let Claude decide." vp-design step 3b will research the domain's visual language and propose domain-unique viz types — at least 2 vizs that could not exist outside this industry. Example: "KPI hero + whatever fits the domain" or "6 vizs, you pick" Recommended: 4-6 for a balanced suite, max 8.

8. Dashboard included?

• Always yes. Every viz pack ships with a Dashboard Studio JSON dashboard that showcases all vizs. This is mandatory — vp-create Step 3c generates it automatically.

Pipeline — FOLLOW THIS EXACT ORDER

The viz pack is NOT done until ALL 4 stages complete. Do NOT skip any stage.

STAGE 1: DESIGN (load vp-design)
  → Brand research, palette, mood, viz inventory
  → Domain visual language step 3b (2+ domain-unique vizs)
  → Output: design brief

STAGE 2: CODE (load vp-viz)
  → Write shared/theme.js, then each viz (formatter + source)
  → Build with build_flat.js after all vizs
  → Output: built viz app directory
  (If format=extension: vp-viz generates config.json + ESM visualization.js instead of formatter.html + AMD source)

STAGE 3: VALIDATE + PACKAGE (load vp-create)
  → Step 2: validate_viz.sh (MUST pass with 0 FAIL)
  → Step 3b: generate_assets.js (icons + gradient bg) + generate_previews.py (per-viz preview.png, 116x76 RGB, Pillow)
  → Step 3c: generate Dashboard Studio JSON with ALL vizs
  → Step 4: package tarball
  → Output: installable .tar.gz
  (If format=extension: vp-create uses yarn build + yarn package instead of build_flat.js + manual tar)

STAGE 4: VERIFY
  → Tarball > 1KB, single top-level dir
  → Dashboard includes all viz types
  → Nav bar points to dashboard

CRITICAL: Stage 3 (vp-create) is where most failures happen. You MUST load the vp-create skill — do NOT package from memory. The dashboard JSON template and composition rules are in vp-create's references.

Write all viz code INLINE (same context). Do NOT dispatch subagents for code generation.

Hand-off to vp-design

When routing to vp-design, pass ALL collected context so it does NOT re-ask:

Design context (from vp-init):
  Format:    {answer to Q1 — "classic" or "extension"}
  Brand:     {answer to Q3}
  Tone:      {answer to Q4}
  Fonts:     {answer to Q5}
  Inventory: {answer to Q7}
  Dashboard: {answer to Q8}

vp-design should verify these 6 fields are present and skip its own Q&A if they are.

Cross-plugin dependencies

Building a viz pack produces 3 artifact types. Each has its own plugin:

Artifact	Plugin	Key skill
Viz source (JS/HTML)	splunk-viz-packs	vp-viz
Dashboard JSON	splunk-dashboard-studio	ds-create
SPL queries	splunk-spl	spl-gotchas

Optional Splunk app dependencies

App	When needed
icon_library	Material Symbols icons in dashboards
infographic_shapes	Gradient shapes, glow effects